Re: [PATCH] riscv: Partially revert "Remove stat64 family from default syscall set"
On Thu, Feb 7, 2019 at 3:04 PM David Abdurachmanov wrote: > > On Thu, Feb 7, 2019 at 11:56 PM Alistair Francis > wrote: > > > > To fix systemd/sysVinit crashes enable __ARCH_WANT_STAT64. > > > > systemd failed to start with this error for 32-bit RISC-V: > > [2.833864] Run /sbin/init as init process > > /sbin/init: error while loading shared libraries: libsystemd-shared-239.so: > > cannot stat shared object: Error 38 > > [2.933593] Kernel panic - not syncing: Attempted to kill init! > > exitcode=0x7f00 > > [2.934120] CPU: 0 PID: 1 Comm: init Not tainted > > 5.0.0-rc4-yoctodev-standard #1 > > [2.934589] Call Trace: > > [2.934919] [] walk_stackframe+0x0/0xa0 > > [2.935243] [] show_stack+0x28/0x32 > > [2.935518] [] dump_stack+0x68/0x88 > > [2.935788] [] panic+0xf0/0x252 > > [2.936041] [] do_exit+0x7de/0x7fc > > [2.936387] [] do_group_exit+0x2a/0x82 > > [2.936674] [] __wake_up_parent+0x0/0x22 > > [2.936982] [] ret_from_syscall+0x0/0xe > > [2.937673] ---[ end Kernel panic - not syncing: Attempted to kill init! > > exitcode=0x7f00 ]--- > > > > sysVinit had a similar problem as well. By enabling __ARCH_WANT_STAT64 > > for 32-bit RISC-V the problem disapears and 32-bit RISC-V is able to > > boot. > > Hi, > > This is expected change for riscv32. More details here: > http://lists.infradead.org/pipermail/linux-riscv/2018-November/002062.html Thanks for the explanation. + Zong Do you know what the statx implementation is in the 32-bit RISC-V glibc submission? Alistair > > david > > > > > Signed-off-by: Alistair Francis > > --- > > This was tested with this fork of glibc to enable 32-bit RISC-V support: > > https://github.com/riscv/riscv-glibc/tree/riscv-glibc-2.29 > > commit: 04fdd476160a55792a75375ba2bf56c761f811c2 > > > > I'm not sure if this is a glibc problem or a kernel problem, but this > > commit caused the breakage between the 4.19 and 4.20 kernel so I'm > > sending out a patch. Let me know your thoughts > > > > arch/riscv/include/uapi/asm/unistd.h | 3 +++ > > 1 file changed, 3 insertions(+) > > > > diff --git a/arch/riscv/include/uapi/asm/unistd.h > > b/arch/riscv/include/uapi/asm/unistd.h > > index 1f3bd3ebbb0d..031b7d78e11c 100644 > > --- a/arch/riscv/include/uapi/asm/unistd.h > > +++ b/arch/riscv/include/uapi/asm/unistd.h > > @@ -20,6 +20,9 @@ > > #endif /* __LP64__ */ > > > > #include > > +#if __BITS_PER_LONG == 32 > > +#define __ARCH_WANT_STAT64 > > +#endif > > > > /* > > * Allows the instruction cache to be flushed from userspace. Despite > > RISC-V > > -- > > 2.20.1 > > > > > > ___ > > linux-riscv mailing list > > linux-ri...@lists.infradead.org > > http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [PATCH] riscv: Partially revert "Remove stat64 family from default syscall set"
On Thu, Feb 7, 2019 at 11:56 PM Alistair Francis wrote: > > To fix systemd/sysVinit crashes enable __ARCH_WANT_STAT64. > > systemd failed to start with this error for 32-bit RISC-V: > [2.833864] Run /sbin/init as init process > /sbin/init: error while loading shared libraries: libsystemd-shared-239.so: > cannot stat shared object: Error 38 > [2.933593] Kernel panic - not syncing: Attempted to kill init! > exitcode=0x7f00 > [2.934120] CPU: 0 PID: 1 Comm: init Not tainted > 5.0.0-rc4-yoctodev-standard #1 > [2.934589] Call Trace: > [2.934919] [] walk_stackframe+0x0/0xa0 > [2.935243] [] show_stack+0x28/0x32 > [2.935518] [] dump_stack+0x68/0x88 > [2.935788] [] panic+0xf0/0x252 > [2.936041] [] do_exit+0x7de/0x7fc > [2.936387] [] do_group_exit+0x2a/0x82 > [2.936674] [] __wake_up_parent+0x0/0x22 > [2.936982] [] ret_from_syscall+0x0/0xe > [2.937673] ---[ end Kernel panic - not syncing: Attempted to kill init! > exitcode=0x7f00 ]--- > > sysVinit had a similar problem as well. By enabling __ARCH_WANT_STAT64 > for 32-bit RISC-V the problem disapears and 32-bit RISC-V is able to > boot. Hi, This is expected change for riscv32. More details here: http://lists.infradead.org/pipermail/linux-riscv/2018-November/002062.html david > > Signed-off-by: Alistair Francis > --- > This was tested with this fork of glibc to enable 32-bit RISC-V support: > https://github.com/riscv/riscv-glibc/tree/riscv-glibc-2.29 > commit: 04fdd476160a55792a75375ba2bf56c761f811c2 > > I'm not sure if this is a glibc problem or a kernel problem, but this > commit caused the breakage between the 4.19 and 4.20 kernel so I'm > sending out a patch. Let me know your thoughts > > arch/riscv/include/uapi/asm/unistd.h | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/arch/riscv/include/uapi/asm/unistd.h > b/arch/riscv/include/uapi/asm/unistd.h > index 1f3bd3ebbb0d..031b7d78e11c 100644 > --- a/arch/riscv/include/uapi/asm/unistd.h > +++ b/arch/riscv/include/uapi/asm/unistd.h > @@ -20,6 +20,9 @@ > #endif /* __LP64__ */ > > #include > +#if __BITS_PER_LONG == 32 > +#define __ARCH_WANT_STAT64 > +#endif > > /* > * Allows the instruction cache to be flushed from userspace. Despite RISC-V > -- > 2.20.1 > > > ___ > linux-riscv mailing list > linux-ri...@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-riscv
[PATCH] riscv: Partially revert "Remove stat64 family from default syscall set"
To fix systemd/sysVinit crashes enable __ARCH_WANT_STAT64. systemd failed to start with this error for 32-bit RISC-V: [2.833864] Run /sbin/init as init process /sbin/init: error while loading shared libraries: libsystemd-shared-239.so: cannot stat shared object: Error 38 [2.933593] Kernel panic - not syncing: Attempted to kill init! exitcode=0x7f00 [2.934120] CPU: 0 PID: 1 Comm: init Not tainted 5.0.0-rc4-yoctodev-standard #1 [2.934589] Call Trace: [2.934919] [] walk_stackframe+0x0/0xa0 [2.935243] [] show_stack+0x28/0x32 [2.935518] [] dump_stack+0x68/0x88 [2.935788] [] panic+0xf0/0x252 [2.936041] [] do_exit+0x7de/0x7fc [2.936387] [] do_group_exit+0x2a/0x82 [2.936674] [] __wake_up_parent+0x0/0x22 [2.936982] [] ret_from_syscall+0x0/0xe [2.937673] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x7f00 ]--- sysVinit had a similar problem as well. By enabling __ARCH_WANT_STAT64 for 32-bit RISC-V the problem disapears and 32-bit RISC-V is able to boot. Signed-off-by: Alistair Francis --- This was tested with this fork of glibc to enable 32-bit RISC-V support: https://github.com/riscv/riscv-glibc/tree/riscv-glibc-2.29 commit: 04fdd476160a55792a75375ba2bf56c761f811c2 I'm not sure if this is a glibc problem or a kernel problem, but this commit caused the breakage between the 4.19 and 4.20 kernel so I'm sending out a patch. Let me know your thoughts arch/riscv/include/uapi/asm/unistd.h | 3 +++ 1 file changed, 3 insertions(+) diff --git a/arch/riscv/include/uapi/asm/unistd.h b/arch/riscv/include/uapi/asm/unistd.h index 1f3bd3ebbb0d..031b7d78e11c 100644 --- a/arch/riscv/include/uapi/asm/unistd.h +++ b/arch/riscv/include/uapi/asm/unistd.h @@ -20,6 +20,9 @@ #endif /* __LP64__ */ #include +#if __BITS_PER_LONG == 32 +#define __ARCH_WANT_STAT64 +#endif /* * Allows the instruction cache to be flushed from userspace. Despite RISC-V -- 2.20.1